Michel Profile picture
Alive! And supporting journalism at the Tarbell Center for AI Journalism. Views my own.
May 22 7 tweets 2 min read
so many AI headlines over the past weeks bear out a simple point: AI companies can't reliably steer their models.

1. Anthropic can't guarantee that Claude 4 won't blackmail users if they make borderline requests

2. OpenAI accidentally made their model way too flattering to users, which they had to roll back.

Apr 28 9 tweets 2 min read
GUYS THE STORY HERE IS NOT ABOUT THE RESEARCHERS IT’S ABOUT THE RESEARCH RESULTS The lack of informed consent is bad.

But these results are crazy Image